World-wide, dates have been set for phase out incandescent bulbs beginning in 2-5 years. White LEDs are the best candidates to fill the need for energy efficient, and it is also hoped cost-effective replacement for the light bulb, with projected costs decreasing from today's over $2 mark to about $0.75 in 2012. A tenfold decrease in the cost of LEDs is required for their wide acceptance for general lighting applications. While the GaN deposition processes and device technology have matured satisfactorily, the low productivity and high costs associated with sapphire substrates used for their deposition are recognized to be a major roadblock to such cost reduction. Replacing silicon wafers for sapphire will solve both these problems. This invention is aimed at solving the hitherto intractable problem of gallium nitride films cracking when grown on silicon.
Today, GaN for LEDs is epitaxially deposited on sapphire (single crystal aluminum oxide), or on single crystal silicon carbide substrates. Production of these substrates is highly specialized, and it is difficult obtain substrates larger than 4′ in diameter, and more than 50% of GaN is grown on 2″ sapphire substrates, and 95% on sapphire substrates range from 2-4″ in diameter. The high cost of the substrates, and low productivity entailed in using small substrates, make GaN LEDs expensive.
In contrast to sapphire and silicon carbide, single crystal silicon wafer are produced in large quantities, relatively inexpensively, and in sizes up to 8-12″ diameter. Using silicon for GaN deposition would greatly reduce the cost of GaN by decreasing the cost of the substrates and, much more significantly, by increasing productivity, in epitaxy and in subsequent processes involved in LED production.
Ability to grow gallium nitride on silicon is a highly sought after objective in the GaN industry. However, two very important technical hurdles have to be overcome to make this a reality. The first of these is of a fundamental nature, viz. that the large lattice mismatch between silicon and GaN will not allow epitaxial growth, i.e. growth of single crystal GaN on single crystal silicon. This challenge also existed for growth on sapphire, and to a lesser extent silicon carbide, where buffer layers of other crystalline materials (Aluminum Nitride, Gallium Aluminum Nitride, etc) are deposited on the substrate to bridge the lattice mismatch. The same approach has been developed for growing GaN epitaxially single crystal silicon substrates. In fact, this has enabled limited commercial production of GaN devices on silicon.
The second hurdle involves the mismatch in the coefficients of thermal expansions (CTEs) of GaN and silicon. The sign and magnitude of this mismatch between them are such that the GaN will be in tension when cooled from the growth temperature of about 1000 C. This tensile stress leads to extensive cracking of the GaN along its easy cleavage planes. A practical effect of this cracking is to limit the thickness and device sizes that can be fabricated. Sapphire is the preferred substrate for GaN thought-out the industry, because its CTE is higher than that of GaN and thereby induces compressive stress in the GaN epitaxial layer when cooled to room temperature. Silicon, on the other hand, puts the GaN in tension as it is cooled to room temperature, and therefore tends to crack for GaN films thicker than few microns.
The coefficients of thermal expansion, CTE, of pure crystalline materials are inherent material properties, and cannot be changed. This problem has therefore not, far lent itself to easy solution so far. Limited relief has been obtained by inserting so-called ‘stress relief’ layers, also deposited epitaxially, on the silicon surface. These surface layers, usually only a few atomic layers thick, require very careful control of the deposition conditions, including by ultra-careful stress monitoring during deposition. In conjunction with such stress control layers, the deposits is either pre-scored, or masked-off with in-situ silicon nitride masking, to limit the lateral dimensions of devices, so that the cracking will not cut though the devices, but led along their boundaries.
In the face of these difficulties, sapphire, as expensive as it is, continues to be the dominant substrate used for GaN LED deposition. It is therefore a need to enable the use of silicon wafers by finding more robust solution to the tensile cracking of gallium nitride.